Skip to content

Command line ​

The paperseek command uses the same application services as the Web UI. Configure model and source credentials in the process environment before a provider-backed search.

Commands ​

CommandPurpose
serveStart the local HTTP server and, by default, a worker
workerRun a separate worker against a local database
searchRun a question from the terminal and return JSON
doctorReport configuration and installation diagnostics
migrateMigrate local data or convert an old configuration
versionPrint the installed version

Use paperseek <command> --help for the installed command's options.

sh
paperseek search "How does open-source funding affect maintenance?" --source openalex --search-budget medium --output json

Supported source IDs: openalex, arxiv, crossref, semanticscholar, pubmed, wos, googlescholar, and paperhub.

Budgets are fast, medium, high, and max. Use --field-id for supported source fields, --field-hint for a textual scope, and --citation-enabled or --no-citation-enabled to select citation expansion where supported.

sh
paperseek search "Efficient long-context language models" --source arxiv --field-id cs.CL --search-budget fast

Model overrides are --llm-provider, --llm-api-type, --llm-model, and --llm-base-url. Read credentials from the environment rather than embedding secrets in command history.

Use a particular database ​

sh
paperseek doctor --database /data/paperseek/paperseek.db --json
paperseek serve --database /data/paperseek/paperseek.db --port 8765

The default is ~/.paperseek/paperseek.db. Share the same path when the Web UI, CLI, and MCP should access the same saved runs.

A diagnostic report helps identify missing configuration; it is not proof that every upstream service will accept a subsequent request. See configuration and migration.

Open source. Open to exploration. Apache-2.0 licensed.